Batch

Run multiple PraisonAI scripts at once for quick debugging and testing

The praisonai batch command discovers and runs all Python files containing PraisonAI imports in a directory. It’s designed for quick debugging and testing of multiple scripts at once.

Basic Usage

bash

# Run all PraisonAI scripts in current directory
praisonai batch

# Include subdirectories
praisonai batch --sub

# Limit recursion depth
praisonai batch --sub --depth 2

Key Features

Auto-discovery: Finds all .py files with from praisonaiagents or from praisonai imports
Server exclusion: Automatically excludes server scripts (uvicorn, Flask, streamlit, etc.) that would hang
Agent filtering: Filter by agent type (Agent, Agents, Workflow)
CI integration: Machine-readable output for automated pipelines
Parallel execution: Run multiple scripts concurrently
Report generation: JSON, Markdown, and CSV reports

Options

Option	Description
`--path, -p`	Path to search (default: current directory)
`--sub, -r`	Include subdirectories
`--depth, -d`	Maximum recursion depth (only with –sub)
`--timeout, -t`	Per-script timeout in seconds (default: 60)
`--parallel`	Run in parallel with async reporting
`--workers, -w`	Max parallel workers (default: 4)
`--server`	Run only server scripts with 10s timeout
`--filter, -f`	Filter by type: ‘agent’, ‘agents’, or ‘workflow’
`--ci`	CI-friendly output (no colors, proper exit codes)
`--quiet, -q`	Minimal output
`--fail-fast, -x`	Stop on first failure
`--include-tests`	Include test files (test_.py, _test.py)
`--no-report`	Skip report generation
`--report-dir`	Custom report directory

Server Script Handling

By default, the batch command excludes server scripts that would hang during execution. Server scripts are detected by patterns like:

uvicorn.run() – ASGI servers
.launch() – PraisonAI agent APIs, Gradio interfaces
app.run() – Flask applications
import streamlit – Streamlit apps
import gradio – Gradio apps
FastAPI(), Flask() – Web frameworks

Running Server Scripts

Use the --server flag to run only server scripts with a 10-second timeout:

bash

# Run only server scripts (10s timeout each)
praisonai batch --server

# Server scripts in subdirectories
praisonai batch --server --sub

This is useful for smoke-testing that server scripts start correctly.

Filtering by Agent Type

Filter scripts by the PraisonAI components they use:

bash

# Only scripts using Agent class
praisonai batch --filter agent

# Only scripts using Agents/PraisonAIAgents (multi-agent)
praisonai batch --filter agents

# Only scripts using Workflow
praisonai batch --filter workflow

CI Integration

The --ci flag provides machine-readable output suitable for CI/CD pipelines:

bash

# CI-friendly output
praisonai batch --ci

# CI with proper exit codes
praisonai batch --ci --fail-fast

CI mode features:

No emoji/color output (plain text)
Proper exit codes (0 = success, 1 = failures, 2 = errors)
Machine-parseable summary

Example CI Output

============================================================
PraisonAI Batch Runner
============================================================
Path: /path/to/scripts
Recursive: False
Timeout: 60s
Scripts: 5
Reports: /Users/user/Downloads/reports/batch/20240115_120000

[1/5] Running: agent_example.py
  PASS PASSED (2.34s)
[2/5] Running: multi_agent.py
  PASS PASSED (5.67s)
[3/5] Running: broken_script.py
  FAIL FAILED (0.12s)
     Error: ImportError: No module named 'missing'

============================================================
SUMMARY
============================================================
  PASSED:  2
  FAILED:  1
  SKIPPED: 0
  TIMEOUT: 0
  -----------------
  TOTAL:   3
============================================================
Reports: /Users/user/Downloads/reports/batch/20240115_120000

Parallel Execution

Run scripts concurrently for faster execution:

bash

# Run in parallel with 4 workers (default)
praisonai batch --parallel

# Custom worker count
praisonai batch --parallel --workers 8

# Parallel with CI output
praisonai batch --parallel --ci

Subcommands

List Scripts

List discovered scripts without running them:

bash

# List all discovered scripts
praisonai batch list

# List with subdirectories
praisonai batch list --sub

# Show available groups only
praisonai batch list --groups

Show Statistics

Display statistics about discovered scripts:

bash

praisonai batch stats
praisonai batch stats --sub

Output includes counts by:

Group (directory)
Runnable status
Agent type (Agent, Agents, Workflow)

View Reports

View the latest execution report:

bash

# Summary view
praisonai batch report

# Show failures only
praisonai batch report --format failures

# Full details
praisonai batch report --format full

# Specific report directory
praisonai batch report --dir ~/Downloads/reports/batch/20240115_120000

Script Directives

Control script behavior with comment directives:

python

# praisonai: skip=true
# praisonai: timeout=120
# praisonai: require_env=OPENAI_API_KEY,ANTHROPIC_API_KEY
# praisonai: xfail=Known issue with API

from praisonaiagents import Agent
# ... rest of script

Directive	Description
`skip=true`	Skip this script
`timeout=N`	Custom timeout in seconds
`require_env=KEY1,KEY2`	Required environment variables
`xfail=reason`	Expected to fail (mark as xfail instead of fail)

Reports

Reports are saved to ~/Downloads/reports/batch/<timestamp>/ by default:

report.json – Full JSON report
report.md – Markdown summary
report.csv – CSV for spreadsheet analysis
logs/ – Individual script output logs

Examples

Basic Testing

bash

# Test all scripts in examples folder
praisonai batch --path ./examples

# Test with subdirectories, 2 levels deep
praisonai batch --path ./examples --sub --depth 2

CI Pipeline

bash

# GitHub Actions / GitLab CI
praisonai batch --ci --fail-fast --timeout 30

# With parallel execution
praisonai batch --ci --parallel --workers 4

Development Workflow

bash

# Quick test of agent scripts only
praisonai batch --filter agent --timeout 30

# Test multi-agent workflows
praisonai batch --filter agents --sub

# Smoke test server scripts
praisonai batch --server

Exit Codes

Code	Meaning
0	All scripts passed
1	One or more scripts failed or timed out
2	Configuration error (invalid path, invalid filter)